I've started a spreadsheet of devices documented to contain Panasonic's video processing chips with TBC. At the moment, that's exclusively the MN6737xx series, but I'd like to broaden it out later. Data sheets are attached for four variants:
Canopus ADVC-300 and Edirol VMC-1 both used MN673744. The latter features analog passthrough I/O, and as I recall there is a block diagram showing that it functions like Panasonic's DVD recorder passthrough (i.e. output has been digitized, DAC'd, and bypasses compression).
In the early 2000s, Canopus Japan made MPEG-2 capture cards with Panasonic chips for line TBC. I bought a used MTVX2004HF, which utilized 2x MN673747 on a daughter board: https://av.watch.impress.co.jp/docs/20040806/dev081.htm
Unfortunately, what I received was semi-functional. Most of the listings I've seen over the years on Yahoo! Japan Auctions simply say that they were pulled from a PC that worked years prior, but I saw one listing which described symptoms that sounded similar enough to mine, even through the garbled Google Translation. This makes me worry that it's a widespread issue. (Bad capacitors?)
That's my starting point for this thread. I'll also cover Analog Devices, Inc. (ADI) chips like ADV7842, but I don't feel like getting into those at this moment.
+ Reply to Thread
Results 1 to 24 of 24
May I add that my Panasonic DMR-E55EG (European model) uses the chip MN85573.
Besides the ADV and Panasonic ICs, some other chips that feature TBC stuff:
- NEC µPD61051, µPD61052 - this is a mpeg2 IC that works with digital input from a separate video ic, but it seems the TBC function is handled here judging by the datasheet. The corresponding video decoder/3D Y/C chips like the UPD64011 doesn't mention it.
The uPD61052 is used in some older dvd-recorders incl Sony RDR-GX7 and Toshiba RD-X2U together with a SAA7118 video decoder, but I'm not sure if the tbc stuff is active in those, seems it required some extra memory.
They seem to later have integrated variants of both into a larger ic together with other things in the EMMA2 (as e.g UPD61171) and EMMA3 architectures, the being used in newer dvd-recorders from pioneer and Sony, and a handful of others like the Toshiba rd-xs24 and various other devices.
In the case of those dvd-recorders, they do correct jitter decently (though not quite as beefy as panasonic ones), though they are a bit more frame drop happy. I don't know if it's doe to configuration or chipset.
NEC Electronics was merged with Renesas at some point, they still seem to make video ICs, but only ones based on techwell and intersil designs (which they also picked up at some point.) judging by their website.
- Toshiba TC90A88F - the TBC chip used in the NV-HS930. While other vcrs seem to use various custom chips, or several discrete ones in older models, this one is an all-in-one off the shelf ting and seems to be used in earlier digital TVs so presumably it wasn't designed to only work in a VCR. Haven't found the datasheet for it (only the simple variant without TBC) but it's listed in the attached toshiba IC pdf.
- Ti TVP5160 - Unlike the very common simplar TVP5150 and TVP514x chips this one states it has "line-level TBC" and has 3D Y/C unlike the other Ti chips, though I have no idea how well it works in practice. Seems to have been used in some TVs and projectors and some very high-end video processor thingies, been discussed a bit on laserdisc forums.
- NEC µPD61051, µPD61052 - this is a mpeg2 IC that works with digital input from a separate video ic, but it seems the TBC function is handled here judging by the datasheet. The corresponding video decoder/3D Y/C chips like the UPD64011 doesn't mention it.
You need to do some more research on the MN6737xx chips. I say this because I've read many documents, but don't remember the specifics off-hand anymore. None of it was good. There are some fatal flaws that make most of the chips unworkable and worthless in most uses. Yes, cards and companies used those chips. But most the devices were infamously crappy as a result. Most never made it outside of Japan or Asia. The Canopus ADVC-300 is a good example of a bad device, as the TBC was weak, so weak as to often do nothing, and when it did actually do something it always did more harm that good. And that specific DV box had other undesired negative image quality issues unrelated to any supposed TBC.
TVP5160 line function is weak, to the point where it just makes the overall image worse. It can't lock anything properly, and it goes to chaos. Any good devices using this chips had to have the line disabled. That's the main issue with the ADV as well, not the the "LLT" is really a TBC -- at least not with defaults, poor docs, nobody has gotten any performance from it to date.
Toshiba DVD recorders were often touted as "TBC replacements", but I just never saw it. Every model I've seen to date was weaker than ES10/15, and that really wasn't a TBC either. Just strong+crippled line, and non-TBC frame sync. If the chips had anything on the Toshiba, being disabled would be my guess as well. You have to realize that stuff on-chip is usually disabled because it didn't work correctly, or at all. Too many people think it's a "hidden feature" but it's not at all the case.
I've been aware of all these for years. Let me save you some time: this endeavor will go nowhere. Been there already. Non-starter.
I'm just sharing information on the technical nitty-gritty
I've never owned the ADVC-300 but I can speak for the Edirol VMC-1, Besides the DV compression it provided rock solid stable images, I've used it with SP, LP, EP, SLP, PAL, NTSC you name it and never failed, Obviously it has a passthrough feature to avoid DV compression but never though about it when I had it and even if I did it wouldn't be a good use for me anyway since I was looking for a device that does TBC and doesn't convert back to analog.
Some discussion of VMC-1 here:
The VMC-1 is a device that gshelley and myself discussed 17+ years ago, some publicly here at VH, some privately. I was actually more interested in the proc amp, 3D Y/C, and AGC. But it never acted the way we wanted it to.
If you insist it's good, I'll re-acquire, re-test, as I have far better/different testing criteria in recent years. But I have to say, the recent MX-1 tests were complete disappointments.
There are a lot of interesting "chips" out there.. and it is helpful to "know" definitely what is inside a particular product.. and its behavior.
I'm not quite of the mindset to (just give up, this will go now where.. don't waste my time).. I think some people, particularly new comers "can" discover new things.
It is rather like poking a bear sometimes tho.. so don't get discouraged.
I've learned to mute topics and posters.. that irritate you sometimes.. you might want to learn up on that too. - you just don't have to interact
As for the MN series.. particularly in the original Japanese.. Matsushita "fan blogs".. the Panasonic silicon did not in the beginning support some features like Wide Screen Aspect Ratio.. or did not include analog front-end Proc-amp features. So the next generation of those chips got those features.
Especially when the very new, first generation of HDTV came out.. 720p.. in Japan they used the D1, D2, ect.. not all chips could do that.
I can't fill in all the gaps.. just be aware of the "feature gaps" between chips
I've been finding out through personal experience that "Surface Mount" electrolytic caps can vary the performance of tbcs and frame sync quite a bit.
They don't bulge or explode their guts all over circuit boards like in switching power supplies.. but can weaken or "attenuate" signals on the input stage of the circuits using the chips. Its something to consider.. before giving up.
I used to ignore the LD-decode effort, until they spawned a new project called VHS-decode which seems to be having some success.
Its not ready for prime time yet.. still at the interesting stage.. but they have switched from using a high priced A to D converter to using old Connextant video capture cards as raw A to D capture devices.. sampling the entire video stream and creating what they label as .tbc files then software processing those to find sync pulses and separating out chroma and luma signals buried in the composite data.
Bob Chi over on the Retro RGB channel interviewed the hardware and software creators on Youtube a couple years ago and they explained how they fell into this LD-decode project.
I'm not into 2D versus 3D Comb filter simulation in pure software, but they seem to have done that.
But for S-VHS vcr's with s-video output they also "seem" to also have been capturing separate chroma.tbc and luma.tbc files to completely by-pass that need for a Comb filter too.
Someday.. it might be a viable alternative to "chips" entirely.. just get a high speed oversampler to "rip" whatever signals can be found and decide how to handle incomplete frames on the fly.
Its "Software Defined" signal processing.
dellsam4 often points people at the Singmai projects for Comb filters, but they also seem to be working on a reclaimed VCR deck that only provides .tbc files for LD-decode or VHS-decode.. angling to provide the most minimal, most crucial pieces of VHS tape digitization before the supply of "working" vcr's become too expensive.. or can no longer be fixed
12-Volt vids often chronicles how fast the video boards and chroma filters in even good quality VCRs still working are failing before the mechanics with no possible repair or replacement.
i speculate (or think).. Singmai might be planning to try and fashion an original transport and tape reader.. to the point of digitizing reading of tape "after" all the VCR's are no more.. but that looks to be a long time from now.. the mechanical parts are still obtainable and they aren't that rare or expensive yet .. the tape binder itself may turn to dust before that is needed
they mentioned something about an autoloader, or mechanical arm to insert and remove tapes.. and I received an inquiry from Arizona about someone wanting to do that
the thing is.. if it will (move) tape, past a helical drum with a working head.. all you need is an A to D capture device
but most of this is just 'fun' stuff for any non-professional
i would go on and on, there are lots of other details about things.. i'm very satisfied with my SDI experiments recently.. its kind of hard to recall what it was like not using SDI now.
AJA has a "mini frame synchronization" device that will perform frame sync decisions on an incoming SDI signal that does not have complete frames and lets you "decide" how to handle those situations.. which is nice.. but expensive.. and its not the FS1 or FS2
i finally found lots of scalers that use the ADV7800 chips with the ADLLT "miniTBC" and "Full frame sync" but not ready to talk about that.. or cringing a bit that people might throw stones at me for bringing it up again
the thing is.. if it doesn't work for you.. because of a missing bit of information.. or misunderstanding.. doesn't mean someone might not figure it out.. or get it working for them.. later
Last edited by jwillis84; 6th Apr 2022 at 15:22.
I hear ya man, once you go SDI you never look back. few years so far and I'm glad I don't have to worry anymore about drivers, Windows OS versions, USB 2.0, capture cards chipsets and versions, chinese knock-offs... I'm so obsessed that one may break down that I have 5 devices now, three BE75 and two S&W TBS800, though I bought them in lots.
Last edited by dellsam34; 6th Apr 2022 at 15:35.
The AverMedia CE310B card that uses Conexant CX23888-322 PCIe A/V Decoder shows TBC-ish behavior in my tests. In my video comparison of the CE310B with the Dazzle DVC 100 you can see skewing/flagging in the output from the Dazzle, but a straight frame in the output from the CE310B. I made several passes, and the results are consistent: the Dazzle shows flagging on an EP recording, while the AverMedia does not.
I did not find any claims regarding TBC functionality in AverMedia or Conexant booklets.
The Conexant CX23888 came out in 2008, pretty late in the SD era.
By then most of the available chips had some answer to the VCR Time base issues and Frame sync issues.
The earlier August 2002 datasheet for the related CX23880 describes the Analog Video Decoder in some detail.
On page 84 in that datasheet it describes the "Ultralock" feature which over samples each line to capture the Overscan and Sync portions of each line.
"For a stable source such as a studio quality source, test signal generator, or DVD player, this variation is
very small (±2 ns). However, for an unstable source such as a VCR, the line length variation is as much as a few microseconds."
Essentially this allows performing Time base correction within the digital domain to filter and re-establish line sync and conformity on the output.
Frame sync is handled using a feature called "Video Sample Rate Converter" separately on page 86 which deviates from the "free running" frame or field output in sync with the input, to "time shifted"output in order to feed a downstream MPEG processor with regular frame or field data as required by that process.
The "Ultralock" for Time base correction
The "Video Sample Rate Converter" for (optional) Frame sync
All makes sense in the 2002 version of the chip, by 2008 it was probably incorporated as a standard component.
I'm not saying the device driver "software" actually made use of these features.. or did it the correct way.. but they do appear to have been aware of why they were included, and were targeting what needed to be handled for hardware MPEG compression.
NEC was sold off to Renesas and stopped producing their competing uPD chips by 2005 and while good, and very high priced Connexant seemed to take up the slack with a much less expensive product.
I would like to remind folks however, that Electrolytic capacitors on the Input can vary (wildly) over time and sabotage the original Time base or Frame sync efforts of what occurs inside these chips.
Last edited by jwillis84; 2nd May 2022 at 15:14.
Thanks, @jwillis84. Right, the CX23888 came out in 2008.
Also, the CX23885 came out in 2007 (a "broadcast audio/video decoder"). This chip is used in the Hauppauge ImpactVCB-e, which lagger uses for his capturing. If it inherits to the CX23880, then maybe it has a similar feature as well.
The datasheet that I found on CX23880 / CX23881 / CX23882 / CX23883 says the following regarding all CX2388x models (CX23885 and CX23888 were not available yet at that time, so who knows how much of the original functionality has been retained):
The CX2388x employs a proprietary technique known as UltraLock to lock to the incoming analog video signal. It always generates the required number of pixels per line from an analog source in which line length can vary by as much as a few microseconds. UltraLock’s digital locking circuitry enables the CX2388x to lock onto video signals quickly and accurately, regardless of their source. Because the technique is completely digital, UltraLock can recognize unstable signals caused by VCR head switches or any other deviation and adapt the locking mechanism to accommodate the source. UltraLock uses nonlinear techniques that are difficult, if not impossible, to implement in genlock systems. And, unlike linear techniques, it adapts the locking mechanism automatically.
UltraLock is based on video signal oversampling using a fixed-frequency, stable clock. Because the video line length varies, the number of samples generated using a fixed-frequency sample clock also varies from line to line. If the number of generated samples per line is always greater than the number of samples per line required by the particular video format, the number of acquired samples can be reduced to fit the required number of pixels per line.
UltraLock accommodates line length variations from nominal in the incoming video by always acquiring more samples than are required by the particular video format and outputting the correct number of pixels per line. UltraLock then interpolates the required number of pixels in a way that maintains the stability of the original image despite variation in the line length of the incoming analog waveform.
Figure 2-41 illustrates an example using 4x Fsc sampling of 3 successive lines of video being decoded for square-pixel NTSC output. The first line is shorter than the nominal NTSC line time interval of 63.5 μs. On this line, a line time of 63.2 μs sampled at 4 × Fsc (14.31831 MHz) generates only 905 pixels. The second line matches the nominal line time of 63.5 ms and provides the expected 910 pixels. Finally, the third line is too long at 63.8 μs within which 913 pixels are generated. In all three cases, UltraLock outputs only 780 pixels.
BTW, the way Ultralock works, and the picture make the discussion of 720 pixels somehow being native and 640 being resampled completely irrelevant, as all of these pixels are resampled from, possibly, unequal number of pixels in each line.
Last edited by ConsumerDV; 2nd May 2022 at 17:30.
The August 2002 datasheet for the related CX23880 also one of the "best" descriptions between the various ways of Separating Y from C I have ever seen.
"Many video decoders employ a luminance notch filter, a chroma band-pass filter, and a chroma comb filter.
The luminance signal is derived by filtering out the color information (chroma) from a composite video signal with a notch filter.
This works because the NTSC color information is in a frequency band centered on 3.58 MHz that extends approximately. +1.3 MHz (i.e., from 2.3 to 4.9 MHz).
The Y filter rejects frequencies in that range.
Although this effectively filters most of the chroma signal out of the luminance signal, it also removes the higher frequency luminance signal components.
This loss of bandwidth reduces the horizontal resolution of the luminance signal, and fine detail in the picture is lost."
Line comb filter
"line comb filters operate by delaying the previous composite video horizontal scan line and comparing it to the current horizontal scan line.
Adding the current line to the delayed line cancels the C signal and provides the Y signal.
Subtracting the current line from the delayed line cancels the Y signal and provides the C signal
This process creates two filters which have a frequency response that look like teeth in a comb.
Which is where the name 'Comb filter' comes from.
This type of filter is usually known as a 1-H comb filter, since it uses one 1-horizontal scan line delay to process the signals..
More complex filters can be built using 2-horizontal scan line delays and are called 2-H comb filters.
While these filters show an improvement to the image versus a notch filter does so at the cost of 50% of the vertical resolution due to the averaging of two lines.
These filters also still suffer the “hanging dot” problem noticeable on the color bar test pattern.
Adaptive comb filter
" To overcome this hanging dot problem and the loss of vertical resolution, a 3-line, adaptive comb filter has been implemented in the CX2388x to
separate the Y/C components in composite NTSC/PAL video signals.
As stated, simple 1-H comb filters can not eliminate “hanging dots” on a vertical color transition.
Comb filtering two successive scan lines with different color values at the same horizontal positions causes the problem.
A line comb filter cannot separate the Y/C signals correctly in this situation.
The color signal crosses over into the luminance signal, creating a cross-luminance artifact.
A 3-line adaptive Y/C separation filter evaluates the video image and selects the most efficient algorithm available in the filter.
This is sometimes called a 2-D filter, because both the horizontal scan lines and vertical transitions are taken into consideration.
This eliminates the hanging dot problem by detecting vertical transitions in the image.
The filter selection logic examines three horizontal scan lines simultaneously.
If a vertical transition occurs between the first and third lines, notch filtering is used instead of comb filtering.
Hence, two lines with different colors will not be input to the comb filter at a transition boundary, eliminating the hanging dot artifact problem.
Which means this filter has two tricks in its toolbox and is 'adaptive' which is where the name comes from.
The CX2388x accomplishes this adaptive processing on a pixel-by-pixel basis, compared to other decoders which can only comb filter on a line by line basis.
Page 94 has a very interesting "Warning" that the Notch filter should never be disabled but could be by accident, and that resolutions below 640x480 cannot use the Comb filter at all due to some issue with the luma sub carrier being outside the range of its abilities.
Failure to manage the notch and comb filter options available properly at various resolutions can result in a monochromatic image entirely on the output.
I found that bit of information very interesting since often with many video capture devices, if you get a B&W picture only at 720x480 switching to 640x480 often restores the color to the image captured.
And some capture devices do not offer a 720x480 capture option at all.. and merely offer 640x480.
It could be that the device driver for those products is not enabling or disabling the notch and comb filter choices available to an adaptive 3 Y/C filter design 'properly' and therefore either restricting resolution to 640x480, or providing 720x480 monochromatic image capture problems.
This datasheet reads much more like a Programmers guide, with registry information and various descriptions of how the chip actually works.. its well worth a reading.
Caution - Y/C filters are (Only) used for Composite signals
A VCR is (not) a native Composite device.
The signal on the VHS tape is recorded in "Color Under" (FM - Frequency Modulated) format which keeps Y separate from C on the tape.
So when its played back it does not have to be recombined into a Composite signal (yellow RCA jack) if the VCR has S-Video output.
S-Video will always keep the Y and C channels isolated and separate.
On the way into a VCR, if it was recorded with a Composite signal, the signal may have had to be "split" by a notch or comb filter before being recorded to tape.. but that is a sunk cost that cannot be reversed.
Whatever the source, the signal on the Tape is in Y/C format and is as good as you can get it off the tape and through an S-Video output.
Don't confuse VHS tape recording with Betamax tape recording, they use different methods, and different resolutions.
Laserdisc used Composite exclusively, and records Composite on the disc, and thus has an entirely different need when considering converting that format into Y/C suitable for video capture cards. In general they want to avoid the "notch" and "comb filters" built-into a capture card.. or want to have at least an "adaptive 3D" filter which chooses from the many different filter "choices" available to it to minimize artifacting. Laserdisc archivists are dealing with no-good-solutions, unlike VCR/VHS archivists.. who have the S-Video option.
It confuses the conversation when someone bounds into a forum thread and starts asking for the "best" Comb filter or any Y/C filter.. most VCR/VHS people would know just avoid Composite at all costs.. stick with S-Video.. and its unnecessary.
Adaptive 2-D vs Adaptive 3D
2-D was taking horizontal and vertical dimensions into account when deciding which filter (notch, comb, ect..) to use on a line or pixel.
3D was taking horizontal and vertical and "temporal" dimensions into account when deciding which filter (notch, comb, ect..) to use on a line, pixel or field or frame
3D was more popular after the advent of MPEG-2 hardware compression, in order to avoid MPEG-2 artifacting due to macroblocks and "noise" perceived from fast motion blur
When capturing 'Uncompressed' adaptive 3D is only relevant when also simultaneously converting from field scans to progressive frame scans.. which smashed two consecutive fields into one frame regardless of the distance in time between the field snapshots.. blurring the motion that takes place between them.
Its a complicated decision to capture Uncompressed, or Compressed as "Progressive frames" only.. it "sounds cool" and ('progressive') but it is ('Always') destructive to the original material being captured.. unless it came from a computer generated game or film media where it started out as a "full frame" image first.
Again... for VCR/VHS with S-Video out.. Y/C filters are a non-issue.. forget them.. put them out of your mind and just avoid Composite signals altogether.
Last edited by jwillis84; 2nd May 2022 at 17:38.
Regarding CX23888, FLP347 posted samples in 2016 with his Osprey 210e that uses this chip (JVC VCR TBC On vs. Off). CX23888 may cope better than the chip in the Dazzle, but likewise, it sucks compared to line TBC.
Last edited by ConsumerDV; 3rd May 2022 at 00:43.
I actually prefer the Dazzle clip here. It's got a slight bend, but it's more constant, instead of warbling like the AVerMedia CE310B.
Here's two fields from AVerMedia as an animated GIF. Deinterlaced in VDub2 and boosted gamma.
Over 20 years
There is always some who prefer the "least" amount of corrections or processing and accept any errors in conversion.
That is a choice.
Some people will sacrifice "Stable" device drivers, for clips of uncommon clarity.
Everyone has their "point of view"
Its an opinion.. and unlike Twitter.. your allowed to have one.
Sadly, The CE310B is not a universal solution: it does not pass Macrovision through, it has issues with A/V sync (I am still finding the right combination of drivers and VDub32/64 versions to capture without a serious de-sync, I was emailing their support, and so far they told me to use older drivers unless I am on Windows 10 despite that their latest drivers are labelled Windows 7/8/10), it shows dot crawl despite they advertise filters that are supposed to clean it up. And overall I don't think the result looks better than from the Dazzle. Basically, the only reason for me to use it is for its TBC-ish behavior. Are you going to add the Conexant chips to your starting post, maybe with some verbiage that it is not super-duper TBC functionality, yet better than nothing?
The Dazzle is going to remain the default ADC for me. It looks softer than the AverMedia, probably because of the composite noise filtering. I saw bending on an LP tape, and flagging on an EP tape. Hollywood movies have looked straight so far
Continuing from the Panasonic DMR settings thread, where we were discussing the sub-variants of PAL models, such as DMR-E55 EB (UK) vs EBL (Ireland).
The DMR-E55PL (PAL-M, Brazil) diagram is a little nicer in that the chip #s are listed on this page, but the PDF formatting is worse.
I've also attached PAL DMR-E65 and DMR-E85 service manuals which use the same chip and feature the same block diagram. In this case, the Region 1 NTSC-only DMR-E85HP / HPC is among the models.
This Russian page has clear photos of the digital PCB from DMR-E65EE: https://www.ixbt.com/dvd/panasonic-dmr-e65ee.shtml
The setup is separate ADC chip -> [3DYC+TBC+FS, split paths to display and to onboard MPEG-2 encoder]. This differs from the previous one based around MN673744 as seen in DMR-E50, for example. There it was [ADC+3DYC+TBC+FS] chip -> split paths to display and to separate MPEG-2 encoder chip.
Thanks Brad for all your effort. Nice to have you back!
Thank you Brad for this info. Good to learn about the circuit architecture and key chips which account for the picture stabilization in passthrough mode for the different models.
Last edited by Sharc; 21st Nov 2022 at 15:09.